The thesis discusses the methods of determination of tumor markers and their clinical importance in medical diagnostics. The theoretical part describes clinically important tumor markers and also the chemiluminescent immunoassay methods used for their determination. The practical part of the thesis describes the introduction of new chemiluminescent methods for the determination of tumor markers in routine operation, evaluates its basic analytical parameters and compares it with the existing immunoturbidimetric method. The practical part also presents results of monitoring of cancer patients with various tumors, which is performed by evaluation of the values of tumor markers. Comparison of our method with other methods within the context of System of external quality control is also documented.

This thesis is focused on the analysis of experimental ECG records drawn up in isolated rabbit hearts and aims to describe changes in EKG caused by ischemia and left ventricular hypertrophy. It consists of a theoretical analysis of the problems in the evaluation of ECG during ischemia and hypertrophy, and describes an experimental ECG recording. Theoretical part is followed by a practical section which describes the method for calculating morphological parameters, followed by ROC analysis to evaluate their suitability for the classification of hypertrophy and at the end is focused on classification.

The diploma thesis is aimed at creating a bankruptcy model for companies of the construction industry in the Czech Republic by using discriminant analysis. In the theoretical part, the concept of bankruptcy model is defined; this part is focused on the inclusion of bankruptcy models in economics, a look into their history, a description of selected models and a brief characteristic of the construction industry. In the practical part, the reliability of selected bankruptcy models is counted and a new bankruptcy model is built.

<p>Uvod: Hronični subduralni hematom (HSDH) je učestala i nezanemarljiva traumatska/netraumatska intrakranijalna lezija, naročito kod pacijenata starije životne dobi. Simptomi pacijenata sa HSDH su raznovrsni i često pogre&scaron;no protumačeni i lečeni. Zbog navedenih činjenica, HSDH predstavlja veliki izazov u dijagnostici i terapiji. Različiti autori ističu značaj radiolo&scaron;kih parametara tokom dijagnostike ovog oboljenja i povezanost sa kliničkom slikom I neurolo&scaron;kim statusom pacijenta, pa je ovo istraživanje bilo usmereno u preciznoj evaluaciji pomenutih parametara, njihovoj prediktivnoj vrednosti i uticaju na prognozu ishoda lečenja. Cilj: Generalni cilj istraživanja je bio da se ispita učestalost pacijenata sa hroničnim subduralnim hematomom na teritoriji Vojvodine u periodu od tri godine; da se analizira starosna dob pacijenata, polna distribucija oboljenja, uticaj komorbiditeta ili faktora rizika na nastanak HSDH; prisustvo ili odsustvo traume koja je doprinela nastanku HSDH, vremenski interval od traume do pojave simpotoma ili znakova bolesti i da se omogući praćenje efekta terapije pacijenata sa ovim oboljenjem. Specifični ciljevi su obuhvatili: 1. Da se utvrde parametri kompjuterizovane tomografije koji imaju prediktivni značaj u pozitivnom ishodu lečenja pacijenata sa hroničnim subduralnim hematomom. 2. Da se utvrde klinički parametri koji imaju prediktivni značaj u pozitivnom ishodu lečenja pacijenata sa hroničnim subduralnim hematomom. 3. Da se dobije model sa najvećom specifično&scaron;ću i senzitivno&scaron;ću za predikciju ishoda lečenja, kombinacijom kliničkih i parametrara kompjuterizovane tomografije kod pacijenata sa hroničnim subduralnim hematomom. Materijal i metode: Istraživanje je obavljeno kao prospektivna trogodi&scaron;nja studija u periodu od aprila 2010. do aprila 2013. godine u Kliničkom Centru Vojvodine- Centru za radiologiju i Klinici za neurohirurgiju i obuvatila je 83 pacijenata sa dijagnozom hroničnog subduralnog hematoma. Svi ispitanici su dijagnostikovani upotrebom kompjuterizovane tomografije glave (CT) i lečeni na Klinici za neurohirurgiju KCV. Izvori podataka su celokupna medicinska dokumentacija svakog pacijenta od perioda prve hospitalizacije do njihovog otpusta, a uključuje i podatke vezane za subjektivni osećaj o zdravstvenom stanju koje su pacijenti usmeno izneli &scaron;est meseci nakon hospitalnog otpusta. Rezultati: Rezultati istraživanja pokazuju da je Glasgow Coma Scala (GCS) tj. nivo svesti pacijenta na hospitalnom prijemu jedini parametar sa visokom prediktivnom vredno&scaron;ću za klinički ishod lečenja pacijenata sa HSDH procenjen preko Glasgow Outcome Scale (GOS). Preostali radiolo&scaron;ki i klinički parametri (&scaron;irina hematoma, pomeraj mediosagitalne linije, denzitet hematoma, starost pacijenta) nemaju visoku prediktivnu vrednost za klinički ishod pacijenata sa hroničnim subduralnim hematomom. Zaključak: Na osnovu grupe analiziranih pacijenata sa HSDH nije bilo moguće napraviti optimalan model za predikciju ishoda lečenja kombinujući radiolo&scaron;ke i kliničke parametre. Pojedinačno posmatrani radiolo&scaron;ki parametri nisu imali visoku prediktivnu vrednost za ishod lečenja pacijenata sa HSDH. Izolovan klinički parametar- GCS- je jedini visoko prediktivni faktor za ishod lečenja pacijenata sa HSDH. Kombinacija kliničkih i radiolo&scaron;kih parametara daje visoku vrednost predviđanja kliničkog ishoda lečenja, ali samo zahvaljujući izrazito visokoj prediktivnoj vrednosti GCS. Iz svega navedenog, kompjuterizovana tomografija (CT) ima veliki značaj u ranoj dijagnostici i praćenju terapije pacijenata sa HSDH, ali CT parametri ponaosob nemaju značaj u predviđanju ishoda lečenja.</p> / <p>Introduction: Chronic subdural hematoma (CSDH) is common traumatic/no traumatic intracranial lesion, especially in older patients. Symptomatology of this disease is variable and often is misdiagnosed and treated with specially challenges in diagnostic and therapy. Different authors pointed on importance of radiological parameters during diagnostic of this disease and connections with clinic and neurological status in patients with chronic subdural hematoma (CSDH), so this thesis was directed to evaluate radiological and clinical parameters of CSDHs and to show their predictive values and their significance on patient&rsquo;s outcome. Aim: General aim of this thesis was to examine frequency of patients with chronic subdural hematoma in Vojvodina, during the period of three years, to analyze the age of population with CSDHs, the gender distribution, an impact of comorbidity or risk factors for patients with CSDHs, the presence or absence of trauma which has contributed to CSDH, to determine time interval from trauma to appearance of symptoms and signs of disease, monitoring the effect of therapy. Specific aims were: 1. To determine clinical parameters with a positive predictive significance on patients outcome 2. To determine radiological parameters with a positive predictive significance on patients outcome 3. To determine optimal prognostic model with high specificity and sensitivity, using combination of radiological and clinical parameters for positive prediction outcome. Material and methods: The study was performed as three-year prospective study from April 2010 to April 2013 in Clinical Centre of Vojvodina, Centre for Radiology and Clinic of Neurosurgery and includes 83 patients with chronic subdural hematoma. All patients were diagnosed using computed tomography of the brain (CT scan) and all were treated in Clinic of Neurosurgery (KCV). Data sources included the medical records of each patient from the time of first hospitalization to period of their discharge and included data related to the subjective feeling of the health that patients verbally present six months after hospital discharge. Results: The results showed that the Glasgow Coma Scale (GCS) - a level of consciousness of the patient on the hospital admission was the only parameter with a high predictive value for clinical outcome of patients with CSDH assessed through Glasgow Outcome Scale (GOS). Other evaluated radiological and clinical parameters (width of the CSDH, mediosagital line displacement, a density of the CSDH, the age of the patient) did not have high predictive values for the clinical outcome in patients with chronic subdural hematoma. Conclusion: Based on the analyzed group of patients with CSDH it was not possible to make optimal predictive model for outcome by combining radiological and clinical parameters. Radiographic parameters did not have high predictive values for treatment outcome in patients with CSDH. Glasgow Coma Scale (GCS) is the only highly predictive factor for treatment outcome in patients with CSDH. The combination of clinical and radiological parameters gives high predictive value for clinical outcome, but only because of extremely high predictive value of GCS. Therefore, computed tomography (CT) is of great importance in early diagnosis and therapy monitoring of patients with CSDH, but CT parameters did not have the high predictive values for the patient&rsquo;s clinical outcome.</p>

Maskininlärning (ML) är idag ett mycket aktuellt, populärt och aktivt forskat område. Därav finns det idag en stor uppsjö av olika avancerade och moderna ML-algoritmer. Svårigheten är att bland dessa identifiera den mest optimala att applicera på ens tillämpningsområde. Algoritmer som bygger på Gradientförstärkning (eng. Gradient Boosting (GB)) har visat sig ha ett väldigt brett spektrum av appliceringsområden, flexibilitet, hög förutsägelseprestanda samt låga tränings- och förutsägelsetider. Huvudsyftet med denna studie är på klassificerings- och regressiondataset utvärdera och belysa prestandaskillnaderna av 5 moderna samt 2 äldre GB-algoritmer. Målet är att avgöra vilken av dessa moderna algoritmer som presterar i genomsnitt bäst utifrån på flera utvärderingsmått. Initialt utfördes en teoretisk förstudie inom det aktuella forskningsområdet. Algoritmerna XGBoost, LightGBM, CatBoost, AdaBoost, SnapBoost, KTBoost, GBDT implementerades på plattformen Google Colab. Där utvärderades dess respektive, tränings- och förutsägelsestid samt prestandamåtten, uppdelat i ROCAUC och Log Loss för klassificering samt R2 och RMSE för regression. Resultaten visade att det generellt var små skillnader mellan dom olika testade algoritmerna. Med undantag för AdaBoost som i allmänhet, med större marginal, hade den sämsta prestandan. Därmed gick det inte i denna jämförelse utse en klar vinnare. Däremot presterade SnapBoost väldigt bra på flera utvärderingsmått. Modellresultaten är generellt sätt väldigt begränsade och bundna till det applicerade datasetet vilket gör att det överlag är väldigt svårt att generalisera det till andra datauppsättningar. Detta speglar sig från resultaten med svårigheten att identifiera ett ML-ramverk som utmärker sig och presterar bra i alla scenarier. / Machine learning (ML) is today a very relevent, popular and actively researched area. As a result, today there exits a large numer of different advanced and modern ML algorithms. The difficulty is to identify among these the most optimal to apply to one’s area of application. Algorithms based on Gradient Boosting (GB) have been shown to have a very wide range of application areas, flexibility, high prediction performance and low training and prediction times. The main purpose of this study is on classification and regression datasets evaluate and illustrate the performance differences of 5 modern and 2 older GB algorithms. The goal is to determine which of these modern algorithms, on average, performs best on the basis of several evaluation metrics. Initially, a theoretical feasibility study was carried out in the current research area. The algorithms XGBoost, LightGBM, CatBoost, AdaBoost, SnapBoost, KTBoost, GBDT were implemented on the Google Colab platform. There, respective training and prediction time as well as the performance metrics were evaluated, divided into ROC-AUC and Log Loss for classification and R2 and RMSE for regression. The results showed that there were generally small differences between the different algorithms tested. With the exception of AdaBoost which in general, by a larger margin, had the worst performance. Thus, it was not possible in this comparison to nominate a clear winner. However, SnapBoost performed very well in several evaluation metrics. The model results are generally very limited and bound to the applied dataset, which makes it generally very difficult to generalize it to other data sets. This is reflected in the results with the difficulty of identifying an ML framework that excels and performs well in all scenarios.

Artificial intelligence techniques, such as artificial neural networks, random forests, or support vector machines, have been used to address a variety of problems in numerous industries. However, in many cases, models have to deal with issues such as imbalanced data or high multi-dimensionality. This thesis implements and compares the performance of support vector machines, random forests, and neural networks for a new bank account fraud detection, a use case defined by imbalanced data and high multi-dimensionality. The neural network achieved both the best AUC-ROC (0.889) and the best average precision (0.192). However, the results of the study indicate that the difference between the models’ performance is not statistically significant to reject the initial hypothesis that assumed equal model performances. / Artificiell intelligens, som artificiella neurala nätverk, random forests eller support vector machines, har använts för att lösa en mängd olika problem inom många branscher. I många fall måste dock modellerna hantera problem som obalanserade data eller hög flerdimensionalitet. Denna avhandling implementerar och jämför prestandan hos support vector machines, random forests och neurala nätverk för att upptäcka bedrägerier med nya bankkonton, ett användningsfall som definieras av obalanserade data och hög flerdimensionalitet. Det neurala nätverket uppnådde både den bästa AUC-ROC (0,889) och den bästa genomsnittliga precisionen (0,192). Resultaten av studien visar dock att skillnaden mellan modellernas prestanda inte är statistiskt signifikant för att förkasta den ursprungliga hypotesen som antog lika modellprestanda.

L'étude de l'évolution du pronostic de santé d'un patient constitue un domaine important en recherche clinique. Récemment, le développement des modèles multi-états a permis d'étudier cette dynamique en prenant en compte plusieurs états de santé. Dans ce manuscrit, nous utilisons plus particulièrement les modèles semi-markoviens. Ce type de processus distingue les temps de séjour dans les états et les trajectoires des transitions, contrairement à l'approche markovienne classique. Nous avons proposé plusieurs adaptations pour pouvoir appliquer ce type de modèle : la censure par intervalle, le choix des distributions des temps d'attente et l'introduction des covariables. Un test d'adéquation est aussi proposé pour vérifier l'hypothèse de stationnarité. Enfin, une méthode originale, incluant la théorie des courbes ROC, est présentée pour définir des états de santé pertinents au regard du pronostic. Ces développements sont principalement appliqués à une cohorte de patients greffés rénaux (base de données DIVAT).

Se recopiló una base de datos geográfica, con cartografía básica y temática, sobre la provincia de Napo (Ecuador), en la que se destacan los mapas de cobertura del suelo de los años 2002 y 2008. Como primer producto se elaboró un mapa de cobertura del suelo del año 1990 a partir de imágenes del sensor TM, (Landsat 4 y 5). Posteriormente se realizó un modelo de probabilidad de presencia de coberturas de tipo antrópico, usando la técnica de regresión logística multivariada; se evaluó el modelo con la curva ROC (Relative Operating Characteristic) y se determinó un alto poder de predicción en el modelo (AUC 0.89), distinguiendo además, que la distancia a centros poblados y a vías de comunicación son las variables más influyentes para la presencia de coberturas antrópicas. Se utilizó el mapa resultante del modelo de probabilidad como entrada en un modelo de transición de coberturas que combina Autómatas Celulares y Cadenas de Markov, entre otros aspectos, simulando un mapa de tipo de coberturas (natural o antrópico) para el año 2008. Se evaluó este mapa simulado, comparándolo con uno de referencia, a partir de índices kappa, y se obtuvo un porcentaje de concordancia general de 93%, lo cual es un buen indicador. Una vez que se ha contado con un modelo que permitía hacer simulaciones con el grado de confianza necesario, se realizaron simulaciones para los años 2015 y 2020. En estos escenarios de tipo de cobertura, se puede ver una clara presión hacia los bosques de la rivera del Rio Napo en un futuro, y también en aquellos cercanos a los principales centros poblados como Tena especialmente. Sin embargo, las áreas protegidas muestran un estado de conservación “natural” en las simulaciones, y esto se debe a sus condiciones de inaccesibilidad, en cuanto a falta de infraestructura vial, y a sus condiciones ambientales especiales. Por último, se verificó que la tasa de deforestación (cambio de natural hacia antrópico) en el período 1990-2008 fue de 4661 ha/año y en el período 2008-2020 sería de 3550 ha/año, indicando que la tendencia en el tiempo muestra en el mejor de los casos una disminución o por lo menos una estabilización de los procesos de deforestación.

Receiver operating characteristic (ROC) curves are widely used for evaluating classifier performance, having been applied to e.g. signal detection, medical diagnostics and safety critical systems. They allow examination of the trade-offs between true and false positive rates as misclassification costs are varied. Examination of the resulting graphs and calcu- lation of the area under the ROC curve (AUC) allows assessment of how well a classifier is able to separate two classes and allows selection of an operating point with full knowledge of the available trade-offs. In this thesis a multi-objective evolutionary algorithm (MOEA) is used to find clas- sifiers whose ROC graph locations are Pareto optimal. The Relevance Vector Machine (RVM) is a state-of-the-art classifier that produces sparse Bayesian models, but is unfor- tunately prone to overfitting. Using the MOEA, hyper-parameters for RVM classifiers are set, optimising them not only in terms of true and false positive rates but also a novel measure of RVM complexity, thus encouraging sparseness, and producing approximations to the Pareto front. Several methods for regularising the RVM during the MOEA train- ing process are examined and their performance evaluated on a number of benchmark datasets demonstrating they possess the capability to avoid overfitting whilst producing performance equivalent to that of the maximum likelihood trained RVM. A common task in bioinformatics is to identify genes associated with various genetic conditions by finding those genes useful for classifying a condition against a baseline. Typ- ically, datasets contain large numbers of gene expressions measured in relatively few sub- jects. As a result of the high dimensionality and sparsity of examples, it can be very easy to find classifiers with near perfect training accuracies but which have poor generalisation capability. Additionally, depending on the condition and treatment involved, evaluation over a range of costs will often be desirable. An MOEA is used to identify genes for clas- sification by simultaneously maximising the area under the ROC curve whilst minimising model complexity. This method is illustrated on a number of well-studied datasets and ap- plied to a recent bioinformatics database resulting from the current InChianti population study. Many classifiers produce “hard”, non-probabilistic classifications and are trained to find a single set of parameters, whose values are inevitably uncertain due to limited available training data. In a Bayesian framework it is possible to ameliorate the effects of this parameter uncertainty by averaging over classifiers weighted by their posterior probabil- ity. Unfortunately, the required posterior probability is not readily computed for hard classifiers. In this thesis an Approximate Bayesian Computation Markov Chain Monte Carlo algorithm is used to sample model parameters for a hard classifier using the AUC as a measure of performance. The ability to produce ROC curves close to the Bayes op- timal ROC curve is demonstrated on a synthetic dataset. Due to the large numbers of sampled parametrisations, averaging over them when rapid classification is needed may be impractical and thus methods for producing sparse weightings are investigated.

INTRODUÇÃO E OBJETIVO: Recidiva da hepatite C crônica com progressão acelerada, embora imprevisível, da fibrose é responsável por piora no prognóstico após o transplante de fígado (Tx). Biópsia hepática protocolar é considerada o padrão ouro para estadiamento da fibrose na recidiva da hepatite C pós-Tx. Para superar as limitações da biópsia, principalmente custo e complicações, marcadores simples e não invasivos de fibrose hepática têm sido propostos para pacientes imunocompetentes, porém com escassos estudos disponíveis no contexto pós-Tx. O objetivo desse estudo é avaliar o desempenho diagnóstico dos marcadores não-invasivos para estadiar fibrose hepática em pacientes pós-Tx. MÉTODOS: Pacientes consecutivos receptores de Tx com recidiva da hepatite C (n=45) que foram submetidos a 118 biópsias hepáticas foram incluídos. Variáveis laboratoriais dentro de trinta dias de cada biópsia foram consideradas. Índice da razão AST-plaqueta (APRI), razão AST/ALT, Escore discriminativo de Bonacini (EDB), Escore de Pohl e índice idade-plaqueta foram calculados para cada biópsia. Fibrose significante foi definida como estágio METAVIR 2. RESULTADO: A área sob a curva ROC (receiver operating characteristic) do Escore discriminativo de Bonacini para predizer fibrose significante foi 0,68, superior aos outros testes avaliados. Utilizando-se o melhor ponto de corte, um valor de Escore discriminativo de Bonacini 8 foi 42% sensível e 95% específico, com razão de verossimilhança positiva e negativa de 7,98 e 0,62, respectivamente. Análise multivariada identificou razão AST/ALT como preditor independente de fibrose significante (OR=4.2; CI 95%=1.5-11.4; p-valor=0.005, ponto de corte 0,89). Análise adicional considerando apenas uma biópsia por paciente confirmou o desempenho superior do Escore discriminativo de Bonacini em relaçãoaos outros testes avaliados, com uma área sob a curva de 0,76. CONCLUSÃO: Escore discriminativo de Bonacini foi o marcador laboratorial não invasivo com melhor desempenho diagnóstico para predizer fibrose hepática significante em pacientes com recidiva de hepatite C crônica pós-Tx / BACKGROUND AND AIM: Recurrent hepatitis C with accelerated, although unpredictable, fibrosis progression accounts for a poor prognosis after liver transplantation (LT). Per protocol liver biopsy is considered the gold standard for fibrosis staging in recurrent hepatitis C after LT to overcome the limitations of liver biopsy, mainly cost and complications, simple non-invasive liver fibrosis tests have been proposed for immunocompetent patients, butfew data are available in the post-transplant setting. The aim of this study was to evaluate diagnostic performance of noninvasive tests to stage liver fibrosis in LT setting. METHOD: Consecutive LT patients with recurrent hepatitis C (n=45) who have undergone 118 liver biopsy were included. Laboratory variables at the time of biopsies were recorded. AST to platelet ratio index (APRI), AST/ALT ratio, Bonacini discriminant score (BDS), Pohl score and age-platelet index were calculated at the time of biopsies. Significant fibrosis was defined as METAVIR stage 2. RESULT: The area under the receiver operating characteristic (ROC) curve (AUC) of Bonacini discriminant score for predicting significant fibrosis was 0,68, better than the other non-invasive liver fibrosis tests. Using the best cutoff value, Bonacini discriminant score value 8 was 42% sensitive and 95% specific, with positive and negative likelihood ratio of 7,98 and 0,62, respectively. Multivariate analysis identified AST/ALT ratio as an independent predictor of significant fibrosis (OR=4.2; CI 95%=1.5-11.4; p-value=0.005, cutoff point 0,89). Additional analysis considering only one biopsy per patient confirmed the superior performance of Bonacini discriminant score compared to the other non-invasive liver fibrosis tests, with an AUC of 0,76. CONCLUSION: Bonacini discriminant score was the non-invasive liver fibrosis test with the best performance for significant liver transplant patients with recurrent hepatitis C